It is easy to spot objects from space. This challenge is more difficult | MarketingwithAnoy

In the spring, when the teams submitted their results to IARPA, evaluator teams judged how well each one did. In June, the teams learned who went on to the second phase of Smart, which runs for 18 months: AFS, BlackSky, Kitware, Systems & Technology Research and Intelligent Automation, which is now part of the defense company Blue Halo.

This time, the teams must make their algorithms usable across different use cases. After all, Cooper points out, “It’s too slow and expensive to design new AI solutions from scratch for any activity we may wish to search for.” Can an algorithm built to find construction now find crop growth? It’s a big shift because it replaces slow-moving, man-made changes with natural, cyclical, environmental ones, he says. And in the third phase, which begins around the beginning of 2024, the remaining competitors will try to turn their work into what Cooper calls “a robust capacity” – something that could detect and monitor both natural and man-made changes.

None of these phrases are strict “elimination rounds” – and there will not necessarily be a single winner. As with similar DARPA programs, IARPA’s goal is to transfer promising technology to intelligence services that can use it in the real world. “IARPA makes phase decisions based on performance in relation to our metrics, diversity of approaches, available resources and the analysis of our independent testing and evaluation,” says Cooper. “At the end of Phase 3, there could be no teams or more than one team left – the best solution could even combine parts from multiple teams. Alternatively, there could be no teams reaching Phase 3.”

IARPA’s investments also often seep beyond the programs themselves and sometimes steer scientific and technological paths, as science goes where the money goes. “Whatever problem IARPA chooses to do, it will get a lot of attention from the research community,” Hoogs said. Smart teams are allowed to continue using the algorithms for civil and civilian purposes, and the datasets that IARPA creates for its applications (such as the tagged satellite images) often become publicly available for other researchers to use.

Satellite technologies are often referred to as “dual-use” because they have military and civilian applications. In Hoogs’ mind, lessons from the Kitware developer software for Smart will be applicable to environmental science. His company already performs environmental science work for organizations such as the National Oceanic and Atmospheric Administration; his team has helped their Marine Fisheries Service detect seals and sea lions in satellite imagery, among other projects. He envisions using Kitware’s Smart software for something that is already a primary use of Landsat images: deforestation. “How much of the rainforest in Brazil has been transformed into man – made areas, cultivated areas?” asks Hoogs.

Auto-interpretation of landscape change has obvious implications for studying climate change, says Bosch Ruiz – for example, seeing where ice melts, corals die, vegetation moves, and soil goes into the desert. Seeing new construction can show where people hit areas of the natural landscape, forest turns into farmland, or farmland gives way to houses.

These environmental applications and their spinout in the scientific world are among the reasons why Smart sought the United States Geological Survey as a test and evaluation partner. But IARPA’s cohort is also interested in the results for their own sake. “Some environmental issues are of great importance to the intelligence community, especially with regard to climate change,” Cooper said. This is an area where the second application of a dual-use technology is largely the same as the first.

Leave a comment